The proinflammatory activity of recombinant serum amyloid A is not shared by the endogenous protein in the circulation.код для вставкиСкачать
ARTHRITIS & RHEUMATISM Vol. 62, No. 6, June 2010, pp 1660–1665 DOI 10.1002/art.27440 © 2010, American College of Rheumatology The Proinflammatory Activity of Recombinant Serum Amyloid A Is Not Shared by the Endogenous Protein in the Circulation Lena Björkman,1 John G. Raynes,2 Chandrabala Shah,2 Anna Karlsson,1 Claes Dahlgren,1 and Johan Bylund1 L-selectin was not down-regulated on RA patient neutrophils as compared with neutrophils from healthy controls. Spiking SAA-rich whole blood samples from RA patients with rSAA, however, resulted in L-selectin shedding. In addition, SAA purified from human plasma was completely devoid of neutrophil- or macrophage-activating capacity. Conclusion. The present findings show that rSAA is proinflammatory but that this activity is not shared by endogenous SAA, either when present in the circulation of RA patients or when purified from plasma during an acute-phase response. Objective. Elevated serum levels of the acutephase protein serum amyloid A (SAA) are a marker for active rheumatoid arthritis (RA), and SAA can also be found in the tissues of patients with active RA. Based on a number of studies with recombinant SAA (rSAA), the protein has been suggested to be a potent proinflammatory mediator that activates human neutrophils, but whether endogenous SAA shares these proinflammatory activities has not been directly addressed. The present study was undertaken to investigate whether SAA in the plasma of patients with RA possesses proinflammatory properties and activates neutrophils in a manner similar to that of the recombinant protein. Methods. Neutrophil activation was monitored by flow cytometry, based on L-selectin shedding from cell surfaces. Whole blood samples from healthy subjects and from RA patients with highly elevated SAA levels were studied before and after stimulation with rSAA as well as purified endogenous SAA. Results. Recombinant SAA potently induced cleavage of L-selectin from neutrophils and in whole blood samples. Despite highly elevated SAA levels, The acute-phase protein serum amyloid A (SAA) is present in the bloodstream at concentrations below 1 M (corresponding to ⬍11 mg/liter) under physiologic conditions, but levels increase significantly during the acute-phase response following infection or conditions of inflammation (1). For example, SAA levels detected in peripheral blood of patients with rheumatoid arthritis (RA) may be well above 50 M (2). In accordance with this, an elevated SAA level in serum is often used as a marker of rheumatic diseases (3). A consequence of long-term elevated levels of SAA is the risk of developing amyloid deposits in the tissues, leading to progressive organ failure (4). In humans, SAA is a family of small apolipoproteins of ⬃12 kd encoded by 4 different genes, of which only 3 are known to be expressed. SAA1 and SAA2 are major acute-phase proteins that are coordinately induced within a few hours during an acute-phase response, and the mature protein sequences share ⬎90% amino acid identity (5). A recombinant SAA1 (rSAA1) has been reported to have cytokine-like properties (6), but most of the literature describing proinflammatory activities has been generated from studies using a commercially avail- Supported by the Swedish Medical Research Council, King Gustaf V’s Memorial Foundation, the Gothenburg Medical Society, the Swedish Society for Medicine, the Inga Britt and Arne Lundberg Research Foundation, and the Swedish State under the LUA/ALF agreement. Dr. Shah’s work was supported by an MRC PhD studentship. 1 Lena Björkman, MD, Anna Karlsson, PhD, Claes Dahlgren, PhD, Johan Bylund, PhD: Sahlgrenska Academy, University of Gothenburg, Gothenburg Sweden; 2John G. Raynes, PhD, Chandrabala Shah, PhD: London School of Hygiene and Tropical Medicine, London, UK. Address correspondence and reprint requests to Lena Björkman, MD, Sahlgrenska Academy, University of Gothenburg, Department of Rheumatology and Inflammation Research, Guldhedsgatan 10A, 413 46 Gothenburg, Sweden. E-mail: lena.i.bjorkman@ vgregion.se. Submitted for publication September 16, 2009; accepted in revised form February 23, 2010. 1660 SERUM AMYLOID A LACKS PROINFLAMMATORY ACTIVITY able recombinant molecule constructed as a mixture of human SAA1 and SAA2. We have, for instance, shown that rSAA activates human neutrophils, and even though the responsible receptor is currently unknown (7), the concentrations of rSAA required for cell activation in vitro (1–10 M) are well below the in vivo levels found in the circulation of RA patients. Since neutrophils are clearly activated by rSAA at low micromolar concentrations, we reasoned that circulating neutrophils from RA patients should display an activated phenotype. We used a sensitive assay that enabled us to monitor neutrophil activation in whole blood and found no evidence of activation, despite high levels of endogenous SAA in the circulation. In contrast, addition of rSAA to blood potently activated neutrophils, regardless of background levels of SAA. In accordance with this, we found that SAA purified from serum during an acutephase response was completely devoid of neutrophilactivating capacity and also lacked any proinflammatory effect when added to adherent peripheral blood mononuclear cells (PBMCs). Our data demonstrate that the SAA present in the circulation during an acute-phase response is functionally different from the recombinant molecule. 1661 pital, using an ELISA kit from BioSource. In parallel, blood samples were collected from healthy volunteers. All blood samples were drawn into heparinized tubes at room temperature and used immediately. Determination of L-selectin expression. The exposure of CD62L (L-selectin) on neutrophils was assessed by immunostaining and fluorescence-activated cell sorter analysis. Briefly, complete heparinized blood (90 l/tube) was incubated with or without stimulation for 20 minutes at 37°C, after which 2 ml of ice-cold FACS Lysing Solution (Becton Dickinson) was added and samples stored on melting ice for 15 minutes. After washing, the cell pellet was incubated with anti–L-selectin antibody (10 l/cell pellet) and analyzed by flow cytometry (FACScan; Becton Dickinson). At least 10,000 events were acquired, and neutrophils were gated on the basis of side and forward scatter. All data were analyzed using WinMDI 2.8 software. TNF production from adherent PBMCs. Human mononuclear cells were isolated by centrifugation over Histopaque 1077, and cells (3 ⫻ 106/ml) were adhered to plastic for 2 hours in serum-free medium. Cells were incubated in RPMI 1640 with 10% fetal calf serum for 6 days before stimulation with control medium, lipopolysaccharide (1 g/ ml), or SAA. Supernatants were collected at 24 hours, centrifuged, and assayed for TNF by ELISA. Statistical analysis. One-way analysis of variance (ANOVA) followed by Dunnett’s multiple comparison test or two-way ANOVA followed by Bonferroni correction was used for statistical evaluation as appropriate. P values less than 0.05 were considered significant. PATIENTS AND METHODS Reagents. Human rSAA (PeproTech) was dissolved in water and stored at ⫺70°C until use; endotoxin levels were ⬍0.1 ng/g (1 EU/g). Endotoxin levels were also tested by standard methods at a clinical bacteriology laboratory (Sahlgrenska University Hospital, Gothenburg, Sweden), and preparations were found to contain ⬍0.056 EU/g. Tumor necrosis factor (TNF) was from Sigma. Phycoerythrin-conjugated monoclonal antibody specific for CD62L (anti–L-selectin) was from BD Biosiences, as was the TNF enzyme-linked immunosorbent assay (ELISA) kit. Endogenous SAA was purified as described previously (8), starting with plasma obtained at plasmapheresis during an acute phase response. Briefly, an SAA ELISA was used to identify fractions containing SAA in sequential stages of hydrophobic interaction chromatography on octyl-Sepharose, gel filtration on Sepharose S-200, and anion exchange on DEAE-cellulose. SAA was ⬎98% pure as determined by sodium dodecyl sulfate–polyacrylamide gel electrophoresis followed by silver staining. On the day before use, samples were dissolved in 5M urea and dialyzed overnight in distilled water, using a 6–8 kd–cutoff dialysis membrane from Spectropor. Patients. After informed consent was provided, peripheral blood was obtained from patients with RA as classified according to the American College of Rheumatology (formerly, the American Rheumatism Association) criteria (9). The levels of endogenous SAA from the patients (n ⫽ 7) were clearly elevated (mean ⫾ SD 450 ⫾ 175 mg/liter [⬃38 M]; range 200–⬎600 mg/liter [17–⬎50 uM]), as analyzed at the clinical immunology laboratory of Sahlgrenska University Hos- RESULTS Recombinant SAA induces activation of human neutrophils in whole blood. We monitored neutrophil activation in whole blood by measuring surface exposure of the cell adhesion molecule L-selectin. L-selectin is present on the surface of resting cells, but is rapidly cleaved off during cell activation. Blood samples from healthy subjects were stimulated with rSAA or with TNF (10 ng/ml) as a positive control, and neutrophils were analyzed by flow cytometry for L-selectin expression (Figure 1A). Whereas unstimulated cells expressed high levels of L-selectin, stimulated blood contained neutrophils with cleaved L-selectin, i.e., with an activated phenotype. Recombinant SAA at the higher concentration used (10 M) induced robust L-selectin shedding (Figure 1B), which was comparable with the level of activation induced by the potent neutrophil activator TNF. Similar results were obtained when neutrophils were purified and washed before stimulation with rSAA in buffer (data available at http://hdl.handle.net/2077/ 22202). These findings show that rSAA activates human neutrophils and potently induces L-selectin cleavage in the presence or absence of blood components, in a concentration-dependent manner. The biphasic appear- 1662 Figure 1. Effect of recombinant serum amyloid A (rSAA) on neutrophil activation in whole blood. A, Blood samples from healthy subjects were stimulated with rSAA (10 M or 0.5 M) or with tumor necrosis factor (TNF; 10 ng/ml) as a positive control, and flow cytometry was performed to assess expression of L-selectin. Unstimulated cells expressed L-selectin in high levels, whereas stimulated blood contained neutrophils with cleaved L-selectin (region 1). B, L-selectin shedding was assessed by flow cytometry. Recombinant SAA (10 M) induced robust shedding of L-selectin, comparable with the level of activation induced by the potent neutrophil activator TNF (ⴱⴱⴱ ⫽ P ⬍ 0.0001). Cleavage of L-selectin by rSAA was concentration dependent; at 0.5 M SAA, no significant cleavage of L-selectin was observed (NS ⫽ not significant). Values are the mean and SD (n ⫽ 3–7 samples per experimental condition). ance of the histogram (Figure 1A) is explained by the fact that very few cells will be intermediary in binding/ fluorescence at any given time. BJÖRKMAN ET AL Neutrophils and monocytes from RA patients with high plasma levels of SAA do not display an activated phenotype. Next we investigated whether high levels of SAA in the circulation of RA patients could activate neutrophils to shed L-selectin. Blood samples were obtained from the 7 patients with elevated serum SAA levels and from healthy controls. Without stimulation, neutrophils from both patients and control subjects were in a resting state (high L-selectin levels) (Figure 2), indicating that endogenous SAA did not activate neutrophils in the circulation. However, when whole blood was stimulated with rSAA, complete shedding of L-selectin from neutrophils was observed (Figure 2). When monocytes from the same samples were analyzed, we found that monocytes from RA patients were likewise not activated, but the addition of rSAA to whole blood led to marked shedding of L-selectin from these cells as well. Resting neutrophils expressed low levels of CR3, and this receptor was up-regulated to the cell surface when the cells were challenged with rSAA or TNF (data available at http://hdl.handle.net/2077/ 22202). Effects were the same in RA patients and controls, indicating that rSAA activates cells in blood regardless of how much endogenous SAA is present. Many proinflammatory effects of SAA have been reported to be neutralized by the presence of highdensity lipoprotein (HDL) or other blood components (5). Our data clearly showed that no blood constituents, either from control or from RA patient blood, were able to neutralize the effect of rSAA. We also performed experiments in which rSAA was incubated with human control plasma (naturally containing HDL) before the cells were stimulated with the mixture. Such preincubation, for up to 2 hours, did not influence the ability of rSAA to induce L-selectin cleavage (data available at http://hdl.handle.net/2077/22202). Purified SAA obtained from serum during an acute-phase response does not activate neutrophils or adherent PBMCs. We then tested whether purified SAA isolated from patients with an acute-phase response had the capacity to activate human neutrophils in a similar manner to that observed with the recombinant protein. Neither of the purified acute-phase isoforms (SAA1 or SAA2) caused significant neutrophil activation when added to blood from healthy subjects (Figure 3A), whereas rSAA was again very effective at inducing shedding of L-selectin from these cells. Such effects were not restricted to selectin expression since other neutrophil responses seen previously with rSAA, e.g., NADPH oxidase activation (7) and the release of interleukin-8 from purified neutrophils (10), were simi- SERUM AMYLOID A LACKS PROINFLAMMATORY ACTIVITY Figure 2. Lack of an activated phenotype in neutrophils from rheumatoid arthritis (RA) patients with high plasma levels of SAA. Blood samples from 7 RA patients with elevated serum SAA levels (mean ⫾ SD 450 ⫾ 175 mg/liter [⬃38 M]) and from healthy controls were analyzed for L-selectin expression by flow cytometry. Neutrophils in unstimulated whole blood from both patients and control subjects were in a resting state (high L-selectin [red]). Stimulation of whole blood with rSAA (dark blue) or TNF (10 ng/ml) (light blue), however, resulted in complete shedding of L-selectin. A, Histograms from a representative experiment. B, Mean and SD results from 7 independent experiments. See Figure 1 for other definitions. larly lacking in experiments using purified SAA (data available at http://hdl.handle.net/2077/22202). 1663 Figure 3. Failure of purified endogenous SAA obtained from serum during an acute-phase response to activate neutrophils or adherent peripheral blood mononuclear cells (PBMCs). A, Neutrophils in whole blood from healthy donors were stimulated with purified endogenous SAA (SAA1 or SAA2; 8 M). Stimulation with SAA1 or SAA2 did not result in shedding of L-selectin from the cell surface, whereas rSAA (8 M) and TNF (10 ng/ml) potently cleaved L-selectin. Values are the mean and SD (n ⫽ 3 samples per experimental condition). ⴱⴱⴱ ⫽ P ⬍ 0.0001. B, Human adherent PBMCs were stimulated with medium, lipopolysaccharide (LPS; 1 g/ml) as a positive control, or purified SAA1 or rSAA at various concentrations. After 24 hours, TNF in supernatants was measured by enzyme-linked immunosorbent assay. Values are the mean ⫾ SD (n ⫽ 6 samples per experimental condition). ⴱ ⫽ P ⬍ 0.05; ⴱⴱ ⫽ P ⬍ 0.01; ⴱⴱⴱ ⫽ P ⬍ 0.001 versus control medium. See Figure 1 for other definitions. 1664 BJÖRKMAN ET AL We also tested the proinflammatory activity of the purified SAA preparations under other experimental conditions in which rSAA has a clear proinflammatory effect. Whereas rSAA effectively induced TNF production from human monocytes, neither SAA1 (Figure 3B) nor SAA2 (data not shown) stimulated release of TNF above background levels. DISCUSSION Increased levels of SAA are frequently observed in the sera of RA patients and have been used as a sensitive marker for disease activity (11). Several studies have established a role for SAA in the pathogenesis of RA as an inflammation mediator, inducing and orchestrating the production of proinflammatory cytokines. SAA has been proposed to be a key regulator of matrix degradation, a hallmark of RA, by causing neoangiogenesis and proliferation of synoviocytes (12–15). Various studies have demonstrated direct receptor-mediated activation of inflammatory leukocytes by rSAA (7,16), even though the question of receptor usage has been a matter of debate. The majority of studies investigating the action of SAA on immune cells have been carried out using a commercially available recombinant form of SAA that is a “consensus SAA molecule.” Human SAA is present in vivo in several different isoforms, and 2 of these, SAA1 and SAA2, are considered proinflammatory variants, the expression levels of which increase immensely during the acute-phase response (5). The amino acid sequence of rSAA corresponds to human SAA1 except for the presence of an N-terminal methionine and substitution of asparagine for aspartic acid at position 60 and arginine for histidine at position 71; these substitutions correspond to the residues found in SAA2, thus forming a hybrid molecule not identical to any of the human isoforms of SAA. As mentioned above, this recombinant chimera has been ascribed numerous different proinflammatory effects and activates various leukocytes at concentrations well below those found in the circulation during inflammatory conditions. We thus reasoned that leukocytes obtained from SAA-rich blood would be activated, and we used peripheral blood from RA patients to investigate this. Cleavage of L-selectin from the cell surface is an early activation marker for neutrophils that proceeds in vivo by the time the phagocytes adhere to the vessel wall and initiate transmigration to the tissues (17). Our data show that in vitro addition of rSAA to blood at concentrations between 0.5 M and 10 M potently activated neutrophils to shed L-selectin. However, neutrophils constantly subjected to up to 50 M of endogenous SAA in the circulation of RA patients were not activated in terms of L-selectin cleavage. An undesired effect of using recombinant (i.e., bacterial-derived) proteins is that proinflammatory effects could be caused by contamination with endotoxin, rather than the recombinant compound itself. For example, the proinflammatory effect of recombinant human C-reactive protein has been shown to be caused by contamination with bacterial products (e.g., endotoxin) and not by C-reactive protein itself (18). The rSAA preparation used in this study contained very low endotoxin levels. Also, rSAA activates neutrophils through a G protein–coupled receptor (7,16) and induces cellular activities (e.g., activation of NADPH oxidase and mobilization of intracellular calcium) not triggered by endotoxin. Nevertheless, we cannot at present entirely rule out the possibility that biologically active bacterial molecules other than endotoxin could accompany the rSAA. Physiologically, it makes sense that neutrophils already in the circulation would not be activated. It is, however, clear that under certain conditions (e.g., sepsis), L-selectin can be cleaved off of neutrophils already in the circulation (Karlsson A, et al: unpublished observations). In addition, stimulation of whole blood (with TNF or rSAA) very clearly activated the cells in this environment, indicating that the shedding machinery is intact and functional. A potential mechanism for avoidance of premature cellular activation by endogenous SAA in the circulation could be that some additional blood component, capable of neutralizing the proinflammatory actions of SAA, is also present in the circulation. Such neutralizing effects have been described for HDL, and SAA is present in plasma mainly as an HDLassociated apolipoprotein (5). However, spiking the already SAA-rich blood with rSAA induced potent neutrophil activation, indicating that HDL or other factors present in the blood failed to neutralize the biologic activity of rSAA. To exclude the possibility that the high levels of endogenous SAA could have saturated the HDL in patient plasma, we incubated rSAA with control plasma (naturally containing HDL) for up to 2 hours, in order to facilitate potential binding to neutralizing blood constituents, without dampening the neutrophil-activating potential. The use of multiple different RA donors and different donor sources of purified SAA rules out differences that may have been due to allelic variants of SAA. Taken together, our findings demonstrate that rSAA differs functionally from SAA found in the circu- SERUM AMYLOID A LACKS PROINFLAMMATORY ACTIVITY lation during inflammatory conditions, and the latter lack neutrophil-activating capacity. The purification process (both for the recombinant as well as for the acute-phase SAA) might render the protein to a nonphysiologic state, although previous work with endogenous SAA, purified or as present in the circulation, has demonstrated its ability to mediate phagocyte recognition of gram-negative bacteria, i.e., SAA functions as an opsonin (19). Even though our data show that the SAA present in the circulation during an acute-phase response lacks direct proinflammatory effects (such as activation of neutrophils), it could still be of functional importance as an opsonin during gram-negative sepsis. It should be noted that SAA is a protein that is prone to aggregation and can assemble in extracellular amyloid plaques in the tissues and have deleterious effects on organ functions (4). The removal of SAA from its carrier HDL may give rise to a form (frequently termed “lipid poor”) that is predisposed to change conformation, potentially in multiple different ways; for instance, a hexameric form has been described (20). For the comparison of SAA preparations used in this study, it was essential that they showed similar aggregation status by gel filtration analysis. The preparations of SAA1, SAA2, and rSAA did exhibit similar aggregation status, providing evidence that aggregation alone is not sufficient to cause activation of neutrophils. The details of SAA aggregation are still largely unelucidated, as is the aggregation state of endogenous SAA found at different sites. Hopefully, future studies will explore whether proinflammatory forms of endogenous SAA exist outside of the circulation, for example, in inflamed RA joints. AUTHOR CONTRIBUTIONS All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Björkman had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study conception and design. Björkman, Raynes, Shah, Karlsson, Dahlgren, Bylund. Acquisition of data. Björkman, Shah, Bylund. Analysis and interpretation of data. Björkman, Raynes, Karlsson, Bylund. REFERENCES 1. O’Hara R, Murphy EP, Whitehead AS, FitzGerald O, Bresnihan B. Acute-phase serum amyloid A production by rheumatoid arthritis synovial tissue. Arthritis Res 2000;2:142–4. 1665 2. Urieli-Shoval S, Linke RP, Matzner Y. Expression and function of serum amyloid A, a major acute-phase protein, in normal and disease states. Curr Opin Hematol 2000;7:64–9. 3. Chambers RE, MacFarlane DG, Whicher JT, Dieppe PA. Serum amyloid-A protein concentration in rheumatoid arthritis and its role in monitoring disease activity. Ann Rheum Dis 1983;42:665–7. 4. Cunnane G. Amyloid precursors and amyloidosis in inflammatory arthritis. Curr Opin Rheumatol 2001;13:67–73. 5. Uhlar CM, Whitehead AS. Serum amyloid A, the major vertebrate acute-phase reactant. Eur J Biochem 1999;265:501–23. 6. Patel H, Fellowes R, Coade S, Woo P. Human serum amyloid A has cytokine-like properties. Scand J Immunol 1998;48:410–8. 7. Bjorkman L, Karlsson J, Karlsson A, Rabiet MJ, Boulay F, Fu H, et al. Serum amyloid A mediates human neutrophil production of reactive oxygen species through a receptor independent of formyl peptide receptor like-1. J Leukoc Biol 2008;83:245–53. 8. Raynes JG, McAdam KP. Purification of serum amyloid A and other high density apolipoproteins by hydrophobic interaction chromatography. Anal Biochem 1988;173:116–24. 9. Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF, Cooper NS, et al. The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum 1988;31:315–24. 10. He R, Sang H, Ye RD. Serum amyloid A induces IL-8 secretion through a G protein-coupled receptor, FPRL1/LXA4R. Blood 2003;101:1572–81. 11. Cunnane G, Grehan S, Geoghegan S, McCormack C, Shields D, Whitehead AS, et al. Serum amyloid A in the assessment of early inflammatory arthritis. J Rheumatol 2000;27:58–63. 12. Migita K, Kawabe Y, Tominaga M, Origuchi T, Aoyagi T, Eguchi K. Serum amyloid A protein induces production of matrix metalloproteinases by human synovial fibroblasts. Lab Invest 1998;78: 535–9. 13. Koga T, Torigoshi T, Motokawa S, Miyashita T, Maeda Y, Nakamura M, et al. Serum amyloid A-induced IL-6 production by rheumatoid synoviocytes. FEBS Lett 2008;582:579–85. 14. Lee MS, Yoo SA, Cho CS, Suh PG, Kim WU, Ryu SH. Serum amyloid A binding to formyl peptide receptor-like 1 induces synovial hyperplasia and angiogenesis. J Immunol 2006;177:5585–94. 15. Mullan RH, Bresnihan B, Golden-Mason L, Markham T, O’Hara R, FitzGerald O, et al. Acute-phase serum amyloid A stimulation of angiogenesis, leukocyte recruitment, and matrix degradation in rheumatoid arthritis through an NF-B–dependent signal transduction pathway. Arthritis Rheum 2006;54:105–14. 16. Su SB, Gong W, Gao JL, Shen W, Murphy PM, Oppenheim JJ, et al. A seven-transmembrane, G protein-coupled receptor, FPRL1, mediates the chemotactic activity of serum amyloid A for human phagocytic cells. J Exp Med 1999;189:395–402. 17. Smalley DM, Ley K. L-selectin: mechanisms and physiological significance of ectodomain cleavage. J Cell Mol Med 2005;9: 255–66. 18. Pepys MB, Hawkins PN, Kahan MC, Tennent GA, Gallimore JR, Graham D, et al. Proinflammatory effects of bacterial recombinant human C-reactive protein are caused by contamination with bacterial products, not by C-reactive protein itself. Circ Res 2005;97:e97–103. 19. Hari-Dass R, Shah C, Meyer DJ, Raynes JG. Serum amyloid A protein binds to outer membrane protein A of gram-negative bacteria. J Biol Chem 2005;280:18562–7. 20. Wang L, Lashuel HA, Walz T, Colon W. Murine apolipoprotein serum amyloid A in solution forms a hexamer containing a central channel. Proc Natl Acad Sci U S A 2002;99:15947–52.